On February 7th, we organized a breakfast in our Paris office to present our collaboration with SNCF Connect & Tech and highlight how the company deployed a server-to-server architecture.


During the conference, Maxence Gama, Digital Project Manager, and François Nollen, Staff Engineer & DevRel at SNCF Connect & Tech, shared their expertise in optimizing data collection with this model.

In this article, we revisit the discussion by exploring the strategic choice of a server-to-server architecture, the technical and ethical challenges related to its deployment, and the assessment presented by the SNCF Connect & Tech and Didomi teams on this project.






The choice of a server-to-server strategy


Launched in 2022 by SNCF Connect & Tech, a private subsidiary of SNCF Voyageurs, SNCF Connect offers an all-in-one digital mobility service (app & website) that allows users to plan their short and long-distance journeys, from itinerary search to booking, customer service, and travel information.

With over 208 million tickets sold in 2023 and an average of 3 million visitors per day, SNCF Connect & Tech aims to provide a simpler, more accessible, greener, and more personalized service. The challenges for the company and its customers are numerous:


  • End of third-party cookies: Improved media algorithms, rapid data activation, more precise targeting, and better attribution.

  • Multichannel tracking and consistency: Multi & cross-channel activation, limited consolidations, and adjustments.

  • Agility and “Time To Market”: Faster deployment (features and fixes) and personalization of customer journeys.

  • Data protection: GDPR compliance, mobile & web consent management, and assurance of adequate data protection.

  • Eco-design and accessibility: Optimization, less dependence on third-party SDKs, accessibility for all customers and devices.

  • Customer experience: More consistent multi-channel experience, fewer crashes, more efficient and faster navigation.


Within this digital transformation, accentuated by the impending end of third-party cookies, SNCF Connect & Tech laid the foundations for robust, secure, and homogeneous data collection by choosing to implement a server-to-server (S2S) architecture to enable more rigorous and respectful data and consent governance.

Here, the operation of a server-to-server application involves setting up a BFF ("Back-for-front") server with which the channels can communicate, then sending the data to a hub that, like a traditional Tag Management System (TMS), allows configuring data processing and dispatching to third-party services.


"All our front channels communicate with our servers. From these servers, we build data layers, doing all the logic of processing, collecting, and tracking, to then contact our partners, with the subtlety of having a hub in between, which will be a visual tool where we can configure things, create rules, everything you already know from a classic TMS."


- François Nollen, Staff Engineer & DevRel at SNCF Connect & Tech


The choice of this S2S-first ("server-to-server first") architecture allows SNCF Connect & Tech to benefit from several advantages specific to the company's use cases:


  • More complete and reliable data: Ability to avoid ad blockers and JavaScript and reduce discrepancies between channels (e.g., fewer discrepancies between web and mobile data, between web and back office).

  • Autonomous digital teams: Teams can deploy various connectors quickly with little technical support (e.g., the ability to run mobile tests in seconds).

  • Uniform data across all channels: Cross-channel logic to offer a consistent and integrated experience (e.g., avoid soliciting the same users via app and web notifications/emails).

  • Better-targeted audience segments: Combining data with granular data to connect attribution tools, the DMP, and the CRM base (e.g., improved targeting and ROI of media campaigns).


These advantages are significant for SNCF Connect & Tech, whose goal is to provide a consistent and respectful data protection experience to users across multiple interaction channels:


"We can address the issues of bots, the web, and the application all in one block, allowing us to have more reliable data and less discrepancy with our back office."


-  Maxence Gama, Digital project manager at SNCF Connect & Tech


Deploying consent management and handling privacy governance


SNCF Connect & Tech took a year to implement this new architecture, starting with the website and application and proceeding to tracking, UX, and user interface.


Didomi x SNCF Connect & Tech - From tags to server-to-server


The implementation also presented new challenges for the SNCF Connect & Tech teams: integrating a new data layer, a hybrid approach to facilitate certain frontend tags, and server-side testing capabilities.

Thus, the data collection and processing set up by SNCF Connect & Tech can be explained in 5 steps:

  1. Front-end collection via user browsing and actions

  2. Local storage on mobile, web, and other channels

  3. Transmission to the server

  4. Backend enrichment and data layer construction

  5. Transmission to the Hub where the data can be processed like in a TMS


Didomi x SNCF Connect & Tech - Architecture server-to-server


Consent management in a server-to-server implementation

Regarding consent management specifically, SNCF uses Didomi's Consent Management Platform (CMP).

The subtlety of SNCF Connect & Tech's integration lies in the company's in-house collection, which, with the help of a Didomi-specific "event listener," is able to synchronize changes made on Didomi to call consent data and transmit it to the server, which then triggers (or does not) partners based on the user's consent status.


Didomi - SNCF Case Study Tech and Connect


The configuration then takes place in the hub through an interface similar to a Tag Management System, where SNCF Connect & Tech teams can make granular decisions regarding data processing.


"The crucial part is to be able to achieve something granular, partner by partner. We sometimes have a choice for some treatments and configuration details: be as precise as possible or simplify processing, depending on needs, and always respect user consent. In other words, there can be several ways to do things from the back office.

The important thing in this case is that it must come down to a user decision or even a business decision, never a choice imposed by the tool or the technical architecture."

- François Nollen, Staff Engineer & DevRel at SNCF Connect & Tech


This type of architecture also offers the possibility of collecting anonymous information when a user refuses consent for analytics, using different and non-reconcilable identifiers to have statistical tracking without ever being able to identify the person or cross the data with other databases.

These principles of pseudonymization and hashing can help maintain measurement capabilities while respecting user choices. They address the most complex scenarios when a user alternates between accepting and refusing consent without ever being able to reconcile the information.

Benefits and limitations of server-to-server data collection


The SNCF Connect & Tech team can assess their server-to-server (S2S) architecture two years after the project's launch. Among the advantages, Maxence Gama and François Nollen identify several positive points for their teams:


  • A robust architecture

  • The business autonomy of a TMS with more serenity

  • An improved “Time-To-Market” allowing teams to deploy features and evolutions every week

  • More reliable and consistent data, which has, for example, reduced discrepancies between the CRM base and analytics from 20% to less than 2% today.


The benefits are also visible to customers:


  • More features, faster and more often

  • The right level of data protection across all channels

  • Reduced customer footprint for more responsible digital practices


Regarding limitations, SNCF Connect & Tech has identified several challenges, looking back over the past two years:

  • A lack of literature available today to learn how to implement an S2S architecture and in-house collection, requiring significant internal efforts

  • Efforts are still necessary to reduce web tags and go forward without certain features or partners.

  • A hybrid architecture induces some complexity, reinforced by major players (ATT, Consent Mode, etc.)

  • More control but also more responsibilities


To conclude, the final assessment is positive for SNCF Connect & Tech, and the event participants testify to general satisfaction with this new architecture.


"The conclusion here is that no one thinks about going back. For us, it has many more advantages than limitations. It may not work for everyone, and it requires efforts and investments, but for us, it's a big step forward."


- François Nollen, Staff Engineer & DevRel at SNCF Connect & Tech


To learn more about consent collection in a server-to-server configuration, schedule a meeting with our team:


Talk to an expert